Research map / Optimization / Optimizers
Training Neural Networks for and by Interpolation
Leonard Berrada, Andrew Zisserman, M. Pawan Kumar
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use ALI-G instead of Adam | Bi-LSTM · SNLI test, cross-entropy | accuracy: 84.5 -> 84.8 | Improved |
| Use ALI-G instead of DFW | Bi-LSTM · SNLI test, SVM | accuracy: 85.2 -> 85.2 | No clear effect |
| Use ALI-G instead of scheduled SGD | WRN 16-4 · SVHN test | accuracy: 98.3 -> 98.1 | Hurt |
| Use ALI-G instead of scheduled SGD | WRN 40-4 · CIFAR-10 test | accuracy: 95.3 -> 95.2 | Hurt |
| Use ALI-G instead of scheduled SGD | DN 40-40 · CIFAR-10 test | accuracy: 95.1 -> 95.0 | Hurt |
| Use ALI-G instead of DFW | WRN 40-4 · CIFAR-100 test | accuracy: 76.0 -> 75.8 | Hurt |
| Use ALI-G instead of scheduled SGD | DN 40-40 · CIFAR-100 test | accuracy: 76.3 -> 76.3 | No clear effect |
| Remove the maximal learning-rate cap (ablation) | ALI-G on Bi-LSTM · SNLI test, cross-entropy | accuracy: 84.8 -> 84.6 | No clear effect |
| Remove the maximal learning-rate cap (ablation) | ALI-G on Bi-LSTM · SNLI test, SVM | accuracy: 85.2 -> 84.7 | No clear effect |
| Replace the capped ALI-G step with an uncapped step (ablation) | ALI-G on an RSI non-convex problem · Non-convex convergence example | convergence | Hurt |
Closest papers by meaning
- A Stochastic Bundle Method for Interpolation (JMLR 2022 · similarity 0.87)
- Painless Stochastic Gradient: Interpolation, Line-Search, and Convergence Rates (NeurIPS 2019 · similarity 0.87)
- Generalization and Stability of Interpolating Neural Networks with Minimal Width (JMLR 2024 · similarity 0.86)
- The Marginal Value of Adaptive Gradient Methods in Machine Learning (NeurIPS 2017 · similarity 0.86)
- Simplicity Bias and Optimization Threshold in Two-Layer ReLU Networks (ICML 2025 · similarity 0.85)
- Global Minimizers of $\ell^p$-Regularized Objectives Yield the Sparsest ReLU Neural Networks (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
